To ensure that your results are valid, you must adhere to the following data requirements.

- Each study part must have a known measurement
- A reference value is the known standard measurement of a part. You use the reference value as a master value for comparison during measurement systems analysis. For example, you use a part that has a known weight of 0.025 g to calibrate your scales.
- Select 8 parts that span the tolerance limit of interest
- To investigate the bias of the gage, select 8 parts for your study that span the tolerance limit of interest and are at approximately equidistant intervals.
- The smallest number of acceptances must equal 0, and the largest number of acceptances must equal 20
- If you specify the lower tolerance limit, the smallest part must have 0 acceptances, and the largest part must have 20 acceptances.
- Each part must be measured by the gage multiple times
- With the AIAG method, you must have exactly 20 trials per part for a full analysis. Minitab performs a partial analysis if you have at least 15 trials; however, Minitab will not provide the bias calculations unless you have 20 trials per part.

- Each reference part must have a known measurement
- A reference value is the known standard measurement of the reference part. You use the reference value as a master value for comparison during measurement systems analysis. For example, you use a reference part that has a known weight of 0.025 g to calibrate your scales.
- Select at least 8 parts that span the tolerance limit of interest
- To investigate the bias of the gage, select at least 8 reference parts that span the tolerance limit of interest and are at approximately equidistant intervals.
- Each part must be measured by the gage multiple times
- With the regression method, you must have at least 15 trials per part.
- The smallest number of acceptances must equal 0, and the largest number of acceptances must equal the number of measurements for the part
- If you specify the lower tolerance limit, the smallest part must have 0 acceptances, and the largest part must have the maximum number of possible acceptances (at least 15 with the regression method).